Broad Learning System Based Model Predictive Control for Nonlinear Industrial Cyber-Physical Systems

This article explores the design of a broad learning system-based model predictive control for nonlinear industrial cyber-physical systems. BLS model is initially trained from data which is obtained by extensive open-loop simulations to accu-rately capture process dynamics. Building on this, a Lyapunov-based MPC framework is designed, employing BLS model as the predictive model. This control method is proved that it can guarantee closed-loop state bounded ness and ultimately convergence to the area around the origin. The experimental section discusses an application in a continuous stirred tank reactors and the results show that the proposed method has efficient control and training time periods.